Texture classification apparatus employing coarsensess and directivity of patterns

• US 5,832,118 A
• Filed: 10/21/1996
• Issued: 11/03/1998
• Est. Priority Date: 05/08/1996
• Status: Expired due to Term

• Current Assignee: Wi-LAN Inc.
• Original Assignee: Daewoo Electronics (Government of South Korea)
• Inventors: Kim, Jin-Hun
• Primary Examiner(s): Mancuso, Joseph
• Assistant Examiner(s): Johnson, Timothy M.
• Application Number:
  US08/734,217
• Time in Patent Office: 743 Days
• Field of Search: 358/261.1, 382/170, 382/168, 382/224, 382/197, 382/194, 382/196, 382/195, 382/203, 382/262, 382/172, 382/245, 382/227, 382/169
• US Class Current: 382/224
• CPC Class Codes: G06T 7/41 based on statistical description of texture

Abstract

An apparatus for classifying a textured image based on pattern coarseness and directivity comprises: a quantizer for obtaining a quantized image from the textured image, the quantized image containing a plurality of pixels each of which is represented by one of N quantized values, N being a positive integer; a scanning block for scanning the quantized image along M scanning directions, M being a positive integer, to thereby provide M scanned images; a grey level mean runlengths configuration block for providing a set of runlengths by counting runlengths of pixels having a same quantized value, for each of the M scanned images and each of the N quantized values, to thereby provide M x N sets of runlengths, providing M×N mean runlengths by averaging each set of runlengths, and forming an M×N matrix whose elements are the mean runlengths; pattern coarseness decision block for determining a coarseness of the textured image by using the matrix; pattern directivity decision block for determining a directivity of the textured image by using the matrix; and texture classification block for classifying the textured image according to the coarseness and the directivity.

Claims

1. An apparatus for classifying a textured image containing a plurality of pixels each of which is represented by a gray level value, which comprises:

quantization means for obtaining a quantized image from the textured image, the quantized image containing a plurality of pixels each of which is represented by one of N quantized values, N being a positive integer;

means for scanning the quantized image along M scanning directions, M being a positive integer, to thereby provide M scanned images;

means for providing a set of runlengths by counting runlengths of pixels having a same quantized value, for each of the M scanned images and each of the N quantized values, to thereby provide M×

N sets of runlengths;

means for providing M×

N mean runlengths by averaging each set of runlengths;

means for forming an M×

N matrix whose elements are the mean runlengths;

means for determining a coarseness of the textured image by using the matrix;

means for determining a directivity of the textured image by using the matrix; and

means for classifying the textured image according to the coarseness and the directivity,

wherein an element of the M×

N matrix is denoted as L_ji, i being a positive integer ranging from 1 to M which designates each of the M scanning directions and j being a positive integer ranging from 1 to N which represents each of the N quantized values and the coarseness is determined as;

##EQU3## K being an integer representing the number of zero elements within the M×

N matrix.

2. The apparatus of claim 1 further comprising a lowpass filter, wherein the textured image is applied to the lowpass filter before it is applied to the quantization means.

3. The apparatus of claim 2, wherein the lowpass filter is a median filter.

4. The apparatus of claim 3, wherein the lowpass filter is a cross shaped median filter.

5. The apparatus of claim 1, wherein the quantization means includes:

means for obtaining a histogram by plotting frequencies of pixels in the textured input image as a function of grey level values;

means for smoothing the histogram;

means for detecting a dynamic range in the histogram, the dynamic range being a range of grey level values between a maximum and a minimum of grey level values whose frequencies are not less than 1 pixel in the histogram;

means for detecting peaks in the histogram wherein a location and a width of each of the peaks are determined;

means for comparing the width of each of the peaks with a peak width threshold; and

means for allotting a quantization interval for each of the peaks whose width is smaller than the peak width threshold, and dividing the remaining dynamic range into (N-P) quantization intervals, P being an integer ranging from 0 to N and representing the number of quantization intervals allotted to the peaks.

6. The apparatus of claim 5, wherein said means for detecting peaks contains:

means for detecting a peak point which is a local maximum of the histogram and whose frequency is larger than an average frequency of the histogram;

means for detecting a left and a right points whose frequencies are smaller than the average frequency and which are nearest to the peak point;

means for determining a left bottom point and a right bottom point which are intersecting points of extensions of lines connecting the peak point with the left point and connecting the peak point with the right point, respectively, with a baseline of the histogram; and

means for determining the width of the peak as the length between the left and the right bottom points.

7. The apparatus of claim 5, wherein the peak width threshold is obtained by dividing the width of the dynamic range by N.

8. The apparatus of claim 1, wherein said M is 4 and said 4 scanning directions are horizontal, vertical, diagonal between an upper right and a lower left corners of the quantized image and diagonal between an upper left and a lower right corners of the quantized image.

9. The apparatus of claim 1, wherein said M is 4, said 4 scanning directions are horizontal, vertical, diagonal between an upper right and a lower left corners of the quantized image and diagonal between an upper left and a lower right corners of the quantized image, the directions corresponds to i=1 to i=4 in that order and the directivity is determined as:

space="preserve" listing-type="equation">Directivity=MAX(D.sub.H, D.sub.V, D.sub.D1, D.sub.D2)

space="preserve" listing-type="equation">wherein;

##EQU4##

10. The apparatus of claim 1, wherein said classification means categorizes the textured image as one of M1 classes according to the coarseness and as one of M2 classes according to the directivity, M1 and M2 being positive integers, to thereby classify the textured image as one of M1×

M2 classes.

11. An apparatus for classifying a textured image containing a plurality of pixels, each pixel being represented by a grey level value, which comprises:

quantization means for obtaining a quantized image by quantizing the textured image based on the distribution characteristic of the pixels in the textured image, wherein the quantized image contains a plurality of pixels represented by one of N quantized values whose quantization intervals have widths adaptively determined based on the distribution characteristic of the pixels in the textured image, N being a positive integer;

means for scanning the quantized image along M scanning directions, M being a positive integer, to thereby provide M scanned images;

means for providing a set of runlengths by counting runlengths of pixels having a same quantized value, for each of the M scanned images and each of the N quantized values, to thereby provide M×

N sets of runlengths;

means for providing M×

N mean runlengths by averaging each set of runlengths;

means for forming an M×

N matrix whose elements are the M×

N mean runlengths;

means for determining a coarseness of the textured image by using the matrix;

means for determining a directivity of the textured image by using the matrix; and

means for classifying the textured image based on the coarseness and the directivity,

wherein an element of the M×

N matrix is denoted as L_ij being a positive integer ranging from 1 to N which represents each of the N quantized values and i being a positive integer ranging from 1 to M which designates each of the M scanning directions and the coarseness is determined as;

##EQU5## K being an integer representing the number zero elements within the M×

N matrix.

12. The apparatus of claim 11, wherein the quantization means includes:

means for obtaining a histogram by plotting frequencies of the pixels in the textured input image as a function of grey level values;

means for smoothing the histogram;

means for detecting a dynamic range in the histogram, the dynamic range being a range of grey level values between a maximum and a minimum of grey level values whose frequencies are not less than one pixel in the histogram;

means for detecting peaks in the histogram wherein a location and a width of each of the peaks are determined;

means for comparing the width of each of the peaks with a peak width threshold; and

means for allotting a quantization interval for each of the peaks whose width is smaller than the peak width threshold, and dividing the remaining dynamic range into (N-P) quantization intervals, P being an integer ranging from 0 to N and representing the number of quantization intervals allotted to the peaks.

13. The apparatus of claim 12, wherein the peaks detecting means contains:

means for detecting a peak point which is a local maximum of the histogram and whose frequency is larger than an average frequency of the histogram;

means for detecting a left and a right points nearest to the peak point whose frequencies are smaller than the average frequency;

means for determining a left bottom point and a right bottom point which are intersecting points of extensions of lines connecting the peak point with the left point and connecting the peak point with the right point, respectively, with a baseline of the histogram; and

means for determining the width of the peak as the length between the left and the right bottom points.

14. The apparatus of claim 13, wherein the peak width threshold is obtained by dividing the width of the dynamic range by N.

15. The apparatus of claim 11, wherein said M is 4, said 4 scanning directions being horizontal, vertical, diagonal between an upper right and a lower left corners of the quantized image and diagonal between an upper left and a lower right corners of the quantized image and the directions corresponding to i=1 to i=4 in that order and the directivity is determined as:

space="preserve" listing-type="equation">Directivity=MAX(D.sub.H, D.sub.V, D.sub.D1, D.sub.D2)

wherein;

##EQU6##